Timeline for Why is Negative Log Likelihood (NLL) a measure of model's calibaration?
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 11, 2021 at 19:37 | comment | added | Lei Huang | I suspect, at a deep level, the statement in the question is related to the fact that maximizing likelihood is asymptotically equivalent to minimizing the KL divergence or cross entropy between the true and predicted distributions. | |
| May 11, 2021 at 19:22 | comment | added | Lei Huang | Okay, edited the answer to use "functional family" instead of "functional class". | |
| May 11, 2021 at 19:22 | history | edited | Lei Huang | CC BY-SA 4.0 |
added 4 characters in body
|
| May 11, 2021 at 19:16 | comment | added | Cagdas Ozgenc | The word “class” is too overloaded in this post. OP may confuse class labels/classification with functional classes. | |
| May 11, 2021 at 16:59 | history | edited | Lei Huang | CC BY-SA 4.0 |
deleted 13 characters in body
|
| May 11, 2021 at 16:52 | history | edited | Lei Huang | CC BY-SA 4.0 |
added 55 characters in body
|
| May 11, 2021 at 16:32 | history | edited | Lei Huang | CC BY-SA 4.0 |
edited body
|
| May 11, 2021 at 16:23 | history | edited | Lei Huang | CC BY-SA 4.0 |
edited body
|
| May 11, 2021 at 16:16 | history | edited | Lei Huang | CC BY-SA 4.0 |
added 5 characters in body
|
| May 11, 2021 at 8:25 | comment | added | Lei Huang | Can you please share the reference to the "calibration - NLL minimization correspondence" statement in your question by the way? | |
| May 11, 2021 at 8:20 | history | answered | Lei Huang | CC BY-SA 4.0 |